Vascular organs-on-chip made with patient-derived endothelial cells: technologies to transform drug discovery and disease modeling

Abstract

Introduction:

Vascular diseases impart a tremendous burden on healthcare systems in the United States and across the world. Efforts to improve therapeutic interventions are hindered by limitations of current experimental models. The integration of patient-derived cells with organ-on-chip (OoC) technology is a promising avenue for preclinical drug screening that improves upon traditional cell culture and animal models.

Areas covered:

The authors review induced pluripotent stem cells (iPSC) and blood outgrowth endothelial cells (BOEC) as two sources for patient-derived endothelial cells (EC). They summarize several studies that leverage patient-derived EC and OoC for precision disease modeling of the vasculature, with a focus on applications for drug discovery. They also highlight the utility of patient-derived EC in other translational endeavors, including ex vivo organogenesis and multi-organ-chip integration.

Expert opinion:

Precision disease modeling continues to mature in the academic space, but end-use by pharmaceutical companies is currently limited. To fully realize their transformative potential, OoC systems must balance their complexity with their ability to integrate with the highly standardized and high-throughput experimentation required for drug discovery and development.

1. Introduction

Drug development is time-consuming and costly, with only a small fraction of drug candidates successfully arriving to market [1]. Despite thorough preclinical investigation and characterization, the majority of therapeutics still fail in clinical trials, often due to safety events or off-target effects [2]. These effects are difficult to predict from the in silico models, in vitro cell culture systems, and small animal studies where the vast majority of drug candidates are first identified and characterized [3]. Despite best efforts to understand disease mechanisms, identify potential toxicities, and otherwise de-risk drugs, some negative effects do not manifest in early studies due to limitations to preclinical model systems and resultingly only present during human clinical trials. Indeed, even drugs that survive the stringent regulatory process and gain approval from the Food and Drug Administration (FDA) have been withdrawn from market due to unforeseen safety issues [4]. In 2016, a systematic literature review of single agent treatments (excluding monoclonal antibody therapies) determined that the second leading cause for withdrawals from markets, accounting for 16% of total withdrawals, was adverse cardiovascular toxicity [5]. Therefore, there is an urgent need to identify cardiovascular risks as early in the drug development process as possible, so efforts and resources can be redistributed to more promising candidates and to protect patients from potential harm.

The frequency of market withdrawal due to unforeseen cardiovascular toxicity can be partially attributed to the fact that key features of the vascular microenvironment are generally absent from 2D culture systems utilized in early drug development. One promising avenue for more robust drug candidate screening in the cardiovascular space involves the utilization of organs-on-chip (OoC). OoC more closely approximate the spatial complexity of native tissues compared to traditional 2D culture systems. Furthermore, OoC can be generated using patient-derived cells, making them attractive for personalized medicine and disease modeling objectives. These systems have gained momentum in recent years, and development is expected to accelerate with the newly adopted FDA Modernization Act 2.0, which removes the regulatory requirement for animal testing for the first time since 1938 [6].

Cardiovascular disease encompasses a broad array of pathologies, ranging from structural cardiac malformations and congenital heart disease to pulmonary arterial hypertension (PAH) and peripheral artery disease [7]. Functionally, the vascular system consists of a closed circuit of vessels that deliver oxygen-rich blood to the periphery and returns oxygen-poor blood to the lungs and heart for reoxygenation and redistribution through the body. At the cellular level, blood vessel structure is heterogeneous and tissue-dependent, but the interior of all vessels is formed from a single-cell thick layer of endothelial cells [8]. Endothelial cells play multiple crucial roles in the vascular system [9]. By signaling to contractile smooth muscle cells, they regulate vascular tone to alter regional blood flow to meet physiological demands [10]. Inability to maintain vascular tone contributes to the progression of several vascular diseases, including hypertension and atherosclerosis [11]. The endothelium also provides a selectively permeable barrier between the circulation and surrounding tissues and regulates the transport of nutrients, waste, and immune cells into and out of tissues [12]. Vascular permeability varies by tissue compartment, with some vessels such as those that invest the kidney being highly fenestrated and others such as the aorta being highly impermeable [13]. The endothelium is also subjected to hemodynamic forces, which govern development and homeostasis [8], and disruption of which contributes to vascular diseases including atherosclerosis and stroke [14]. Thus, the development of preclinical models that better recapitulate the native endothelial microenvironment has focused on recapitulating endothelial cells in environments with relevant ECM mechanical properties, physiological flow, and 3D architecture.

2. Approaches for the design and fabrication of vascular organs-on-chip

Broadly speaking, OoC couple miniature biomimetic tissues (“organs”) within microfluidic devices (“chips”) to generate physiologically relevant culture systems at a small scale (Figure 1) [15]. While traditional 2D cultures typically fail to replicate the complex 3D architecture and dynamics of the native vasculature, OoC enable the manipulation of key features of the tissue microenvironment such as geometry, extracellular matrix mechanical properties, and fluid dynamics, making them attractive tools for the study of vascular disease. The considerable advancements in the fabrication and implementation of OoC for disease modeling and drug discovery have been extensively reviewed elsewhere [15–23], and recent work has demonstrated that vascular OoC recapitulate and are predictive of endothelial cell signaling and mechanisms in murine blood and lymphatic vasculature [24,25]. The enabling microfluidic devices for OoC have become increasingly standardized and commercialized, and the fabrication methods and tools are increasingly accessible to the broader biomedical research community. Thus, the key design and fabrication challenges are the biological components of OoC, such as the extracellular matrix and cells.

Figure 1. Methods for fabrication of OoC.

(A) Pulmonary-artery-on-chip utilized by Ainscough et al to evaluate the role of the endothelium in PAH (Reproduced from [49]) (B) Representative device (Reproduced from [69]) used to generate microfabricated vessels to evaluate transport barrier function (Reproduced from [29]). (C) Microvascular networks formed by HHT patient-derived cells in a microfluidic device Reproduced from [71]. (D) An engineered vessel integrated with a peristaltic pump originally described by Fernandez et al (Reproduced from [70]). This device was later utilized to investigate the endothelial contribution to HGPS by Atchison et al [68]. Panels used with permission. PAH=pulmonary arterial hypertension. HHT=hereditary hemorrhagic telangiectasia. HGPS=Huntington-Gilford Progeria.

Several key approaches have been developed to construct vascular OoC [26–29], and the devices implemented for disease modeling or drug screening can be broadly classified in 4 categories (Figure 1, Table 1): A) Microfabricated substrates – microfluidic devices in which cells are cultured on substrates such as polydimethyl siloxane (PDMS) patterned by microfabrication techniques; B) Patterned hydrogels – microfluidic devices or wells in which a 3D hydrogel is cast in a manner that creates cylindrical voids within the hydrogel to serve as a template for cell seeding; C) Self assembled vascular networks – microfluidic devices that contain endothelial cell-laden hydrogels in which the endothelial cells coalesce to form perfusable, interconnected vascular networks through a process that resembles developmental vasculogenesis; D) Tissue engineered blood vessels – multi-layered vessels that integrate self-assembly and patterning to create perfusable vessels that resemble arterial vasculature. Each approach has advantages and drawbacks for recapitulating certain disease characteristics and for fabrication throughput, which are discussed below.

Table 1. Summary of donor-derived endothelial cells and their use in OoC for disease modeling.

Endothelial cells used includes donor-derived endothelial cells only and excludes commercially acquired endothelial cells.

Disease	Cells	EC source	Verification of endothelial identity	Vascular modeling strategy	Diameter of native vasculature	Wall shear stress	Method of perfusion	ECM coating/hydrogel	Ref
PAH	EC + SMC	BOEC, healthy (n = 5 donors) BOEC, PAH (n = 5 donors)	Marker expression by IF (VE-cadherin)	Microfabricated substrate: 2 channels separated by a 400nm PET membrane within PDMS. Each channel was 200um high × 1000um wide	~25mm	6 dyn/cm2	Peristaltic pump	Rat tail collagen I coated PDMS and PET	49
HHT	EC + PC	iPSC, healthy and HHT (n = 1 mosaic donor)	Marker expression by IF (CD31, VE-cadherin, vWF, VEGFR2); Functional assay (electrical resistance for barrier function)	Self assembled vascular networks: Single channel device with parallel media channels, 1.3mm wide × 250um high	~5–100um* *Genetic disease, affects all vessels. Particularly affects regions where arteries feed into venous circulation	Static	NA	Fibrin hydrogel	71
Type I diabetes (in pigs), thromboinflammation	EC	BOEC, healthy (n = unpublished)	Marker expression by IF (ICAM1, vWF, VE-cadherin)	Patterned hydrogel: Single channel device in PDMS. Channel was 200 μm wide × 75 μm high	>300um	0.81 dyn/cm2 overnight to induce alignment after seeding devices, 25 dyn/cm2 when perfusing devices with whole blood	Syringe pump	Rat tail collagen I coated PDMS	82
Peripheral artery disease	EC + MSC	BOEC, hospitalized volunteers (n = 10 donors)	Marker expression by IF (CD31, vWF)	Decellularized ECM scaffold	~2mm	Static	NA	Decellularized bovine carotid artery	85
Type I diabetes	EC + porcine islets	BOEC, healthy (n = unpublished)	Marker expression by FACS (CD31, VE-cadherin, VEGFR2); Marker expression by IF (CD31, VEGFR2, VE-cadherin, vWF) Functional assay (tube formation)	Decellularized ECM scaffold; self assembled vascular networks	>300um	Static	NA	Decellularized rat lung GFR Matrigel	84
PAH	EC	BOEC, healthy (n = 3 donors) BOEC, PAH (n = 3 donors)	Marker expression by FACS (CD133, CD34, VEGFR2); Gene array analysis; Functional assay (tube formation)	Self assembled vascular networks	~25mm	Static	NA	NA	81, 30, 31
HGPS	EC + SMC	iPSC, healthy (n = 2 donors) iPSC, HGPS (n = 2 donors)	Marker expression by IF (CD31, VE-cadherin, vWF, VEGFR2) Marker expression by FACS (CD31, VE-cadherin) Marker expression by Western blot (CD31, VE-cadherin)	TEBV: iSMC-laden collagen gelled around 810um steel mandrel and plasticly compressed to generate lumen that was seeded with iEC to generate vessel	~4mm* *Genetic disease, affects all somatic cells.	6.8 dyn/cm2	Peristaltic pump	Rat tail collagen I hydrogel	68
DCM	EC + CM	VEC, DCM (n = 1 donor) BOEC, DCM (n = 1 donor) iPSC, DCM (n = 7 donors) iPSC, healthy (n = 2 donors)	Marker expression by qPCR (NRP1, EFNB2, NOTCH1) Marker expression by FACS (CD31) Marker expression by qPCR (CD31, eNOS) Functional assays (tube formation, NO production)	Self assembled vascular networks	~4mm* *Genetic disease, affects all somatic cells. Researchers measured endothelial dysfunction in brachial arteries.	15 dyn/cm2	Ibidi Pump System	GFR Matrigel	78

PAH=pulmonary arterial hypertension. HHT=hereditary hemorrhagic telangiectasia. HGPS=Hutchinson-Gilford Progeria syndrome. DCM=dilated cardiomyopathy. iPSC=induced pluripotent stem cell. BOEC=blood outgrowth endothelial cell. VEC=vessel endothelial cell derived from surgical resection of vessel. SMC=smooth muscle cell. iCM=induced cardiomyocyte. iEC=induced endothelial cell. iSMC=induced smooth muscle cell. rMSC=rat mesenchymal stromal cells. hMSC= human mesenchymal stromal cells. PC=pericyte. IF=immunofluorescence. FACS=fluorescence activated cell sorting. qPCR=quantitative polymerase chain reaction. PDMS=polydimethylsiloxane. TEBV=tissue engineered blood vessel. NA=not applicable.

3. Scope of literature review

Here, we focus on critically reviewing and comparing the cells that have been used in vascular OoC toward informing the next generation of disease models and systems for preclinical drug candidate screening. We highlight studies where patient-specific vascular models were used to identify novel disease features or interrogate responses to therapy. While the number of such studies remain limited, the approaches and insights provided by careful review will inform future development of vascular OoC and study design utilizing OoC. We first summarize two studies illustrating the utility of patient-specific iPSC for vascular disease modeling, followed by a brief summary on blood outgrowth endothelial cells (BOEC) as an alternative approach to patient-specific vascular disease models. We briefly highlight other applications for patient-specific endothelial cells in disease modeling, therapeutic screening, and drug development before closing with an outlook on the future of OoC in drug discovery and development.

4. Cell types for disease modeling in OoC

The structure and composition of blood vessels vary along the vascular tree [8]. The inner lining of all vessels consists of a single-cell thick layer of endothelial cells that are in direct contact with blood, while the composition of the vessel wall extending from the basolateral endothelial surface varies among subtypes of the vascular tree and is tissue-specific. For example, capillaries are comprised of lumenized endothelial cells with intermittent pericytes, while arteries possess stratified layers of smooth muscle cells, fibroblasts, and extracellular matrix. Thus, the focus of the development of vascular OoC has been on recapitulating the endothelial microenvironment toward establishing functional barrier between blood and surrounding tissue, but increasingly vascular support cells including smooth muscle cells (SMCs), pericytes, and fibroblasts are included in OoC disease models (Table 1).

SMCs, which surround larger vessels, are contractile and drive vasodilation and vasoconstriction in response to signals released by the endothelium [32–34]. SMCs play an important role in a variety of vascular diseases including atherosclerosis and hypertension, usually through the development of aberrant proliferative or contractile phenotypes [35–37]. Therefore, their inclusion in vascular OoCs is crucial to investigate complex interplay between cell types present in native vessels. Similarly, pericytes are typically found along smaller vessels and capillaries [38]. These cells are particularly important in the vasculature of the central nervous system, where they contribute to capillary development in the brain and help to maintain vessel stability and barrier function [39,40]. Fibroblasts in larger vessels play an important role in secreting extracellular matrix proteins that help to maintain structural integrity [41,42]. Thus, an important design consideration for vascular OoC, that in part determines the method of fabrication (Figure 1), is the structure and composition of the native vessels that drive the disease, and the sourcing of endothelial cells and relevant support cells (Table 1).

5. Limitations in the use of commercially available primary cells for disease modeling

A central challenge in the study of vascular disease and in screening for novel therapeutics is the availability of appropriate endothelial cells. Commercially available primary endothelial cells, such as the widely used human umbilical vein endothelial cell (HUVEC), have several limitations. In the case of HUVEC, the tissue compartment of origin presents a potential drawback, as endothelial phenotypes and behavior can vary with the tissue compartment from which they are derived [43]. Beyond tissue niche, endothelial cell behavior has also been shown to vary with the age [44] and sex of donor [45]. Thus HUVEC, which are derived from perinatal tissues, might fail to recapitulate certain features of adult or diseased endothelial biology [46]. Similar drawbacks arise from the use of other common commercially available endothelial cells, such as human lung microvascular cells (HMVEC-L) or human aortic endothelial cells (HAEC), whose native physiological environments are distinct. For instance, the hemodynamic forces to which HAEC are exposed differ greatly from those that HMVEC-L might experience in vivo [47]. Another barrier to the study of vascular disease is the recapitulation of underlying genetics in disease models. In some cases, a driving pathogenic mutation has been identified and can be genetically introduced to a population of cells to study disease mechanisms and potential therapeutics [48]. More often, for example in diseases such as PAH, a complex combination of genetic alterations contributes to disease pathogenesis, making the generation of appropriate cell models difficult and time consuming [49]. Furthermore, for most diseases the driving genetics or causal factors are unknown, and the use of healthy cells could provide misleading or inaccurate preclinical data.

6. iPSC-derived endothelial cells

Cellular reprogramming allows for terminally differentiated cell types to be reprogrammed into induced pluripotent stem cells (iPSC) and then re-differentiated into cell types of the vasculature, including endothelial cells [50], cardiomyocytes [51], and vascular smooth muscle cells [52]. Induced cardiomyocytes in particular have been used to generate mature cardiac tissues ex vivo as well as in the generation of linked OoC systems used to characterize drug pharmacokinetics, demonstrating the utility of reprogramming for the study of cardiovascular disease and toxicity [53,54]. Additionally, transdifferentiation from one adult cell type to another is possible and has been successfully performed to generate smooth muscle cells from endothelial progenitor cells to generate OoC with genetically matched vascular cells [55–57]. The methods for induction of pluripotency in somatic cells have been extensively reviewed elsewhere [58], and are promising for generating endothelial cells for the study of disease [59]. By using easily harvested cells such as fibroblasts or peripheral blood mononuclear cells (PBMC) to derive other less accessible cell types such as endothelial cells, patient-specific cells that are otherwise inaccessible or challenging to maintain in culture can be generated (Figure 2). Because they maintain their original genotypes through the reprogramming process, these cells overcome the disease modeling challenges presented by the use of primary cells as discussed above. iPSC-derived vascular cells can then be integrated with OoC to study underlying disease mechanisms or to screen therapeutics. These approaches are particularly promising for the study of diseases driven by complex genetic alterations, where generation of accurate cell or animal models can be challenging and time consuming.

Figure 2. Generation of patient-derived vascular cells for disease modeling.

Patient-derived cells can be obtained through reprogramming strategies, which allow for generation of multiple cell types, or by culturing PBMC until BOEC emerge. These cells can then be integrated with OoC for precision disease modeling. PBMC=peripheral blood mononuclear cells. BOEC=blood outgrowth endothelial cells. OoC=organ on chip. Created with BioRender.com

In vivo, endothelial cell progenitors develop from multipotent mesodermal cells early during embryogenesis. Differentiation into the endothelial fate is guided by signaling molecules such as bone morphogenetic protein 4 (BMP4), fibroblast growth factor (FGF), and vascular endothelial growth factor (VEGF) among others [60]. At the transcriptional level, members of the E26 transformation-specific (ETS) and Forkhead box (FOX) families of transcription factors regulate endothelial specification [61]. In fact, ETV2 is a transcription factor that has shown promise for direct reprogramming into endothelial cells [62,63]. Methods for EC differentiation from iPSC typically mimic some parts of this process. For example, common differentiation protocols involve the administration of specific cytokines and growth factors during culture, such as VEGF and FGF, ultimately directing iPSC into an endothelial fate over the course of around 10 days [64]. iPSC-derived EC (iEC) are typically benchmarked by their expression of common endothelial cell markers, such as von Willebrand factor (vWF) or vascular endothelial cadherin (VE-cadherin), or by comparing their functional activity to other known endothelial cells [65]. Endothelial cells are highly heterogeneous in vivo, and it is unlikely that one marker panel or assay would suffice to define their maturity [9]. Still, iEC offer a promising alternative to other broadly used EC such as HUVEC which are typically difficult to genetically modify and begin to lose regeneration potential over the course of serial passaging [66]. Recently, iEC from patient-derived iPSC have been utilized within microfluidic devices to study complex genetic diseases that would be challenging to model in commercially available cells such as HUVEC.

7. Disease modeling and drug screening in OoC with iPSC-derived endothelial cells

Hutchinson-Gilford progeria syndrome (HGPS) is a disease caused by a point mutation in the gene coding for lamin A that is typically associated with detrimental effects on smooth muscle cells (SMC) [67]. A novel role for the endothelium in HGPS was recently revealed by utilizing cells generated from patient-derived iPSC to create a 3D engineered microvessel disease model [68]. While HGPS has long been associated with cardiovascular disease, early studies using 2D culture systems failed to identify a role for EC in HGPS clinical progression. In this study, Atchison et al [68] demonstrated that engineered vessels fabricated using iEC and iSMC from HGPS patients displayed dysfunctional flow responses and altered vasoconstrictive and vasodilatory behaviors (Figure 1D). By further generating vessels that contained various combinations of healthy or diseased iEC and iSMC, vessel dysfunction was demonstrated to be independent of iSMC disease state, thereby elucidating a role for the endothelium in HGPS. By identifying a novel cell type that contributes to disease pathogenesis, this study motivates the investigation of new therapeutic strategies targeting the endothelium in HGPS and provides a model for the characterization of other potential drug candidates that might already be in development.

The use of prefabricated structures to generate tissue geometry is typically referred to as “top-down” engineering and can be used to generate endothelial vessels-on-chip (Figure 1B) [68–70]. Another approach, typically referred to as “bottom-up” engineering, creates an environment in which cells can self-assemble into native tissues (Figure 1C). The ability of endothelial cells to form microvascular networks in 3D can serve as a readout of their functionality, and microfluidic chips are well-suited to house these networks. In 2022, Orlova et al [71] successfully used iPSC derived from a patient with a rare mosaic form of HHT to study the effects of endoglin mutation, the driving mutation of HHT, on EC in this manner. Patient fibroblasts from skin biopsy were reprogrammed into iPSC using episomal vectors and subsequently differentiated into EC using defined media. Due to the mosaic nature of the patient’s tissues, both endoglin mutant and wildtype isogenic endothelial cells were generated from the iPSC, thus generating an ideal isogenic control for the study. To investigate the differences between the endoglin mutant iEC and wildtype iEC, iEC and human brain vascular pericytes were embedded in fibrin hydrogels and allowed to form 3D microvascular networks (Figure 1B). Using this experimental system, features of endothelial dysfunction such as increased permeability and topological network disorganization were revealed in microvessels generated from mutant iEC that did not present in microvessels generated from wildtype vessels or in 2D culture models. HHT, which manifests through tortuous vessels and arteriovenous malformations (AVM), is currently incurable and therapeutic intervention is limited to surgery [72]. Research into the underlying cellular and molecular mechanisms is challenged by limitations of available mouse models, where suspected compensatory and adaptive mechanisms dampen vascular defects in vivo. By using patient-specific EC in an appropriate microphysiological setting, Orlova et al [71] demonstrated that key features of HHT can be recapitulated in an OoC.

Together these studies demonstrate two examples where vascular disease modeling was enabled by integrating iPSC technology with OoC models. Specifically, the results illustrate the utility of these technologies for modeling of vascular diseases that are predicated by complex genetic alterations and are otherwise difficult to recapitulate in cell or animal models.

8. Blood outgrowth endothelial cells from peripheral blood draw

Blood outgrowth endothelial cells (BOEC) are derived from endothelial colony-forming cells (ECFC) that exist in peripheral circulation. Here, we use the term BOEC to refer to as the in vitro derivatives of ECFC isolated from circulation [73]. In vivo, ECFC are typically recruited to sites of vascular injury to repair damaged vessels [74]. Importantly, these cells are distinct from mature circulating endothelial cells (CEC), which typically enter circulation after being shed from sites of vascular injury [75]. Derivation of BOEC is relatively straightforward and involves only the culture of PBMC on collagen-coated substrates in endothelial growth media supplemented with extra fetal bovine serum (Figure 2) [76]. Over the course of one to two weeks with routine media changes, BOEC will begin to adhere and proliferate. Once colonies form, they can be expanded and cultured like commercially available endothelial cells. BOEC offer several advantages over other primary endothelial cells or iEC, especially in their proliferative potential, which is typically higher than that of iEC [77]. They are also less technically demanding to produce than iEC, requiring only collagen-coated substrates and media changes for their derivation as opposed to more complex reprogramming regimes, while still resulting in endothelial cells with genotypes identical to the donor. However, yield can vary among donors which can present challenges to researchers. Studies report different rates of success in BOEC isolation, ranging from 0 to 75% and varying with age and disease state [75]. Nevertheless, due to their accessibility via minimally invasive procedures such as peripheral blood draw and the relatively low barriers for derivation, BOEC have shown increasing utility in the study of vascular diseases in recent years.

9. Disease modeling and drug screening in OoC with BOEC

In 2020, Sayed et al [78] used a complementary modeling approach incorporating both iEC and BOEC derived from patients or healthy donors to investigate the impact of lovastatin on patients suffering from LMNA-related cardiomyopathy. LMNA-related cardiomyopathy is a disease driven by mutations in lamin A, causing cardiovascular defects that can culminate in sudden cardiac death or end-stage heart failure [79]. Though lamin A is expressed in endothelial cells, little was known about the contribution of the endothelium to the pathology of LMNA-related cardiomyopathy. Using RNA-seq and ATAC-seq to compare healthy and diseased iEC, the authors identified KLF2 as a differentially down-regulated gene in diseased cells. These findings were then corroborated by genetic manipulation of patient-derived iPSC using TALEN-based genome editing. Specifically, lamin A mutation was corrected in diseased iPSC and introduced to healthy iPSC prior to their differentiation into iEC, thus generating ideal isogenic controls to validate hits identified in -omics screens. Because KLF2 expression was reduced in diseased cells as well as healthy cells carrying the TALEN-mediated LMNA mutation, a focused panel of 16 small molecules capable of regulating KLF2 was selected for screening. This targeted screen revealed the therapeutic potential of the drug lovastatin, which rescued defects in tube formation assays and improved nitric oxide production in vitro. Motivated by the possibility of repurposing lovastatin, which is already FDA-approved, patients were then treated with lovastatin over the course of 6 months. Clinical features of the disease were improved in patients after lovastatin treatment. These clinical observations were further confirmed at the cellular level by comparing BOEC collected from a patient before and after lovastatin treatment. BOEC collected from patients after treatment had reduced functional defects compared to BOEC collected before. This study represents the utility of matched ex vivo models for targeted drug screening and direct translation.

BOEC have also been used multiple times to generate matched models of pulmonary arterial hypertension (PAH). PAH is a disease which typically develops under a complex set of conditions, which can include microenvironmental factors such as hypoxia or inflammation as well as genetic alterations in genes such as bone morphogenetic protein type II receptor (BMPR2). These features combined with its slow progression make PAH a difficult disease to model in cells or animals [80]. In 2022, Ainscough et al [49] generated a biomimetic model of PAH incorporating BOEC isolated from healthy donors or patients and commercially sourced SMC. This pulmonary artery-on-chip featured an interface between a layer of human pulmonary SMC and donor-derived BOEC that closely mimics native arterial architecture. The chip was also integrated with a peristaltic pump, allowing for the application of fluid shear stress (Figure 1A). Using this system, two clinically relevant PAH drugs were screened for their ability to attenuate SMC proliferation, a key feature of PAH progression. Both demonstrated efficacy in the system, suggesting that this PAH model could serve as a viable platform for drug screening. In another study, BOEC from healthy or diseased donors were used to identify a novel therapeutic target for treatment of PAH [81]. In this work, Macias et al [81] screened a novel HIF2a inhibitor for efficacy using BOEC as a model. They performed genetic knockdown manipulations in patient BOEC and utilized a tube formation assay to evaluate endothelial function in the presence of the inhibitor. Their results suggested that inhibition of HIF2a was successful in ameliorating endothelial hyperproliferation in vitro, motivating further study of the novel inhibitor as a potential drug candidate. Still others have utilized BOEC to model diabetes-related thromboinflammation by culturing BOEC from healthy volunteers in flow-conditioned vessel chips and exposing them to diabetes-related inflammatory cytokines [82]. Together, these studies demonstrate that the use of patient derived primary cells in OoC provide a route for novel drug target discovery.

10. The endothelium in tissue-specific OoC models

Beyond modeling diseases specific to the cardiovascular system, endothelial cells play an important role in other disease models and in tissue engineering. As engineered organs become larger and more complex, passive diffusion is no longer sufficient to meet metabolic demand. Recapitulating native vasculogenesis and angiogenesis ex vivo for vascularization of OoC is a promising strategy to overcome some of the challenges that accompany ex vivo tissue generation and to increase the physiological relevance of the resulting tissue [83]. One example recently published by Citro et al [84] demonstrated the applicability of healthy BOEC in the generation of a xenogeneic vascularized transplantable organ. In this study, BOEC from healthy volunteers were introduced into decellularized extracellular matrix from rat lung, where they autonomously coalesced into a vascular network and successfully integrated with neonatal porcine islets (NPI) to generate a vascularized endocrine pancreas (VEP). This vascular bed was perfusable, suggesting that it was mature and functional. Further, NPI that were cultured in vascularized lung scaffolds produced more insulin and glucagon than NPI non-vascularized scaffolds, suggesting that their functionality was improved with the addition of a supporting vascular network. Importantly, these VEP were immediately productive upon grafting into diabetic mice, and 37% of mice required no insulin post-transplantation. By using BOEC to vascularize the VEP and demonstrating its potential for grafting, this work highlights the potential for patient-matched grafting of VEP. Other groups have also utilized decellularized matrices as scaffolds for generation of patient-specific tissue engineered vascular grafts (TEVG). For example, Seiffert et al [85] demonstrated that ECFC isolated from hospitalized patients could be used to generate small diameter TEVG as a potential strategy for intervention in peripheral artery disease, a disease which typically requires surgical replacement of affected vessels. Incorporation of vascular cells has also been demonstrated to improve functional outputs of cardiac organoids [86] and to enhance neuronal development on chip [87].

Recent efforts seek to integrate distinct OoC systems via engineered vessels or endothelialized channels in order to recapitulate organ-to-organ cross talk and to better screen for systemic toxicities. So-called “human-on-chip” or “multi-organ chip” systems have been used to model various physiological networks, including the blood-brain barrier (BBB) [88] as well as to link liver spheroids to intestinal or skin cultures [89]. One recent study linked iPSC-derived heart, bone, skin, and liver OoC with vascular flow and validated the system by identifying biomarkers of doxorubicin cardiotoxicity, thus generating a potential model for patient-based drug safety screening [90]. Because the endothelial barrier plays a vital role in pharmacokinetics in vivo, incorporating endothelialized linkages between existing OoC is likely to improve their relevance as substitutes for other animal models used to screen for toxicity [91]. OoC systems that include a vascular compartment or vascular interface are likely to be better poised to identify these potential toxicities.

11. Conclusion

The vasculature is as crucial as it is complex. Cardiovascular diseases are one of the leading causes of death worldwide, and there is urgent need for better therapeutic interventions [92]. Additionally, due to the near omnipresence of the vasculature in the body and the role of the endothelium in pharmacokinetics, inclusion of vascular tissue compartments in disease models is likely to improve their predictive power. Because of interspecies differences between humans and classic animal models, there is motivation to generate human OoC from patient-derived cells [3]. These OoC would eliminate interspecies differences inherent to animal models while also enabling personalized medicine endeavors.

The two strategies for generation of patient-derived endothelial cells described in this review have unique advantages and disadvantages. Cellular reprogramming allows for the generation of multiple cell types from one starting pool of somatic cells, whereas collecting BOEC from peripheral blood does not. Additionally, due to donor-to-donor variability, ECFC may exist in lower numbers in certain individuals and thus reduce the chance of successfully generating BOEC colonies [30]. On the other hand, compared to standard reprogramming strategies, collection of BOEC requires less technical expertise, effort, and time. The approaches are not mutually exclusive and can be used in a complimentary fashion to great effect, as demonstrated by Sayed et al [78].

Here, we summarize a selection of examples where donor-derived endothelial cells were utilized for translational research (summarized in Table 1). These cells have demonstrated their utility in disease modeling [49,68,71], in vascularization of transplantable organs [84,85], in the linkage of OoC systems for toxicity screening [89,90], in targeted drug screens [49,78,81], and more. As indicated by research published in the last few years alone, patient-specific endothelial cells are likely to play a pivotal role in the advancement of disease modeling and personalized medicine.

12. Expert opinion:

The drug development process is long and complex. Recent advances in computational chemistry, machine learning and artificial intelligence have improved in silico screening capabilities, thus dramatically increasing the number of promising compounds that might feasibly enter a preclinical development pipeline [93]. Advanced in silico screening methods are also being deployed for drug repurposing, a burgeoning field which seeks to apply existing drugs to novel indications, thus accelerating the delivery of treatments to patients [94]. With these advancements and others, there is no shortage of promising drug candidates. Increasingly, the bottleneck in early drug development is experimental validation of potential drug candidates in models that are predictive of toxicity and efficacy in humans. High throughput screening methods for preclinical characterization seek to overcome this bottleneck, often allowing thousands of molecules to be iteratively screened across multiple cell lines in 2D assays to identify leads for further investigation. Unfortunately, this throughput usually comes at the cost of physiological relevance, with few of the drug candidates that show promise in early 2D or 3D culture systems progressing to clinical trials.

The longstanding requirement for animal testing in the regulatory process sought to complement preclinical in vitro data with in vivo screening, thus producing safety and efficacy profiles for drugs that were presumably more relevant to humans. A lack of concordance between animal and human drug responses has motivated a shift in this paradigm, however, with many funding agencies and regulatory bodies steering away from animal models in recent years. Disagreements between animal models and human physiology are common and are particularly prevalent in the cardiovascular space, where they have been well documented [95–97]. In response to these observations, the FDA Modernization Act 2.0 was signed into law in the United States in December 2022 [98]. For the first time since 1938, the requirement for animal testing in the drug approval process has been removed, provided that suitable alternatives can be utilized. This law acknowledges advancements OoC technologies and seeks to streamline and improve the American regulatory process. The United States is not alone in this endeavor. In 2021, a resolution to phase out animal testing was approved by European Parliament, reflecting a broader trend towards alternative testing methods [6]. These developments offer an exciting opportunity for the continued development and improvement of OoC for disease modeling and drug discovery.

Despite these regulatory and legal efforts, a lack of standardization and the current bespoke nature of OoC development are significant barriers to adoption by private sector pharmaceutical companies that lead clinical development. These companies have invested heavily in preclinical research and development (R&D) infrastructure and significant revision to established pipelines will be costly [99]. Furthermore, while a growing number of biotechnology companies have been founded to subcontract OoC development, the lack of standardization and throughput, along with lack of standardized protocols and quality control metrics, has prevented broader adoption by the contract research organizations that increasingly perform preclinical testing for companies leading clinical development [99]. However, due to the low rates of commercial success for compounds with preclinical promise as determined by current methods, some recent estimates suggest a total cost reduction of up to 26% with OoC implementation during preclinical R&D [100].

Recently, efforts have been undertaken to establish unified language and standards to enable streamlined development and coordination between academic laboratories and industry [101]. For example, in the US, the Standards Coordinating Body recently published Standard Terminology Relating to Microphysiological Systems (American Society for Testing Materials Standard F350–22) and similarly, the European Organ-on-Chip Society (EUROoCS) seeks to implement a roadmap for broader platform adoption and development through coordination between academic and commercial entities. Such standards not only address gaps in understanding and communication between academic researchers and commercial end-users, but also will allow better communication of industry needs to academic labs and foundries. There is a long-standing tradition of academic foundries that offer technology, services, and training to commercial entities for micro- and nanofabrication such as the Research Triangle Nanotechnology Network (RTNN) and the National Nanotechnology Coordinated Infrastructure (NNCI). Convergence on broad fabrication standards could enable similar infrastructure for development, fabrication, commercialization, and use of OoC and allow for the establishment of shared foundries staffed by process and technology experts to alleviate the lack of resources at smaller institutions or companies.

The studies reviewed here demonstrate the immense promise of OoC in disease modeling and drug screening, as each study identified novel drug targets that were not revealed by more traditional models (Table 1). Yet, each study was performed in a different device, with different cell types, sourced using different protocols (Table 1). While the custom nature of these approaches is feasible in academic research, this model of development remains too costly to justify even the charitable estimate of a 26% cost reduction in R&D expenses. As summarized in this review, we identify cell source as a significant barrier to standardization that could enable the economy of scale to drive broader adoption of OoC toward revolutionizing preclinical models. While the patient- and disease-specific nature of primary BOEC and iPSC remain the central advantage, the current lack of standardization in protocols for derivation and quality control remain preclusive to broader adoption. Multiple funding agencies have programs dedicated to supporting the continued development of OoC systems, including the National Institute of Health’s Tissue Chip for Drug Screening Program and the Defense Advanced Research Projects Agency’s Microphysiological Systems Program [102]. In the private sector, pharma companies such as GlaxoSmithKline and Johnson & Johnson have partnered with biotech companies to support the development and optimization of their OoC platforms [102]. As development of OoCs continues in both the academic and private sectors, balancing throughput and complexity will be a key consideration for OoC technology implementation and disease modeling in general (Figure 3). Overall, however, recent studies such as the ones described in this review indicate that patient-specific disease models can be successfully leveraged for targeted drug screens, and that human OoC are capable of identifying toxicities not predicted by animal models.

Figure 3. Complexity in vascular disease modeling.

Disease modeling aims to mimic the disease phenotype that presents in a sick patient ex vivo, and various factors contribute to overall model complexity. Increased system complexity recapitulates patient physiology at the expense of throughput and ease of use. Several studies discussed in this review are presented based on the complexity of the engineered system vs. complexity of the biological samples. iEC=induced endothelial cell. iCM=induced cardiomyocyte. DCM=dilated cardiomyopathy. HHT=hereditary hemorrhagic telangiectasia. iSMC=induced smooth muscle cell. HGPS=Huntington-Gilford Progeria syndrome. BOEC=blood outgrowth endothelial cell. PAH=pulmonary arterial hypertension. Created with BioRender.com

Article highlights:

• Animal models suffer from interspecies differences in cardiovascular biology and physiology, and adverse cardiovascular events are the second leading cause for market withdrawal of approved drugs.

• New United States federal law and guidance from the Food and Drug Administration (FDA) incentivize and in some cases require the utilization of alternatives to animal models for marketing authorization.

• Organs-on-chip (OoC) provide functional models built from human cells and have potential for improved predictive power in disease modeling and toxicity screening.

• An increasing number of vascular OoC have been developed to identify interventional drug targets for various diseases.

• Patient- and disease-specific OoC have been generated with donor-derived vascular cells generated via cellular reprogramming or by collection from peripheral blood, and each approach has technical and scientific benefits and drawbacks.

• Vascular OoC hold immense promise in identifying novel disease mechanisms and interventional drug targets, as demonstrated the studies herein, with the aim of working toward the development of patient-specific OoC for drug discovery and screening.